🤣🤣🤣 I’m in tears. Actual tears.

I’d love to see them introducing a control group for their “toxicity” model by including subs from their other identified clusters. How can you know what it means for tankies to be millions of billions toxic if you don’t have baselines?

Ironically(?) the funding is to develop a machine learning algorithm not to spot and moderate racism but to spot and moderate the least racist of any two examples. Which means, the project is to develop a comparative model but they haven’t thought about using comparison within the research itself. Meanwhile, real scholars get fired from all over the place for being in unions and demanding a living wage.